photo-identification software for bowhead whale images
photo-identification software for bowhead whale images
photo-identification software for bowhead whale images
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
SC/60/BRG24.<br />
PHOTO-IDENTIFICATION SOFTWARE FOR BOWHEAD WHALE<br />
IMAGES<br />
GILBERT R. HILLMAN * , KELLY D. TRASK + , KATHRYN L. SWEENEY # , ANDREW R. DAVIS + , WILLIAM R. KOSKI + ,<br />
JULIE MOCKLIN # AND DAVID J. RUGH #<br />
* 5030 Marathon Drive, Madison, WI, 53705, USA.<br />
+ LGL Limited, environmental research associates, 22 Fisher St., P.O. Box 280, King City, Ontario, L7B 1A6,<br />
Canada.<br />
# National Marine Mammal Laboratory, Alaska Fisheries Science Center, National Marine Fisheries Service,<br />
National Oceanic and Atmospheric Administration, 7600 Sand Point Way, NE, Seattle, WA 98115-0070, USA.<br />
ABSTRACT<br />
A computer program has been developed <strong>for</strong> <strong>identification</strong> of individual <strong>bowhead</strong> <strong>whale</strong>s in aerial <strong>photo</strong>graphs. The program<br />
enables the user to create a collection of <strong>images</strong> and to compare an image to the collection. The collection is ranked by the<br />
program based on similarity to the query image, facilitating <strong>identification</strong> of the individual in the image. The program has been<br />
tested by the developer (GRH) with 20 <strong>images</strong> and it appears to compute its matching functions correctly. Eleven successive<br />
versions of the program were provided to LGL Limited, who tested the program with a collection of 78 <strong>images</strong> and have provided<br />
reports on bugs in the system and feature requests to the developer. All known bugs have been fixed and many features have been<br />
added to make the program per<strong>for</strong>m better. Additional modifications to the program are likely to improve matching per<strong>for</strong>mance<br />
and will be identified when larger numbers of <strong>images</strong> have been processed through the program.<br />
KEY WORDS: BOWHEAD WHALE, PHOTO-ID, MARK-RECAPTURE, ARCTIC, COMPUTER-MATCHING<br />
INTRODUCTION<br />
The <strong>photo</strong>graphic database of the Bering-Chukchi-Beau<strong>for</strong>t Seas (BCBS) population of <strong>bowhead</strong> <strong>whale</strong>s<br />
(Balaena mysticetus) contains almost 18,000 <strong>images</strong> collected from the late 1970s to 2007. More than half of the<br />
14,682 <strong>images</strong> that have been classified according to image quality are of adequate quality to determine if the<br />
<strong>whale</strong>s are marked and could be recognized if they were <strong>photo</strong>graphed in the future. All of the <strong>images</strong> in the<br />
database, but particularly the high quality <strong>images</strong>, have provided useful in<strong>for</strong>mation on life-history parameters<br />
including calving intervals (Miller et al., 1992; Rugh et al., 1992a), growth rates (Koski et al., 1992, 1993),<br />
population structure (Davis et al., 1983; Koski et al., 1993, 2006, 2007; Angliss et al., 1995), population size<br />
(Rugh, 1990; da Silva et al., 2000; Schweder, 2003; Schweder and Sadykova, 2007; Koski et al., 2008) and<br />
survival rates (Whitcher et al., 1996; Zeh et al., 2002). These data continue to be used in various on-going<br />
studies to further describe various parameters of <strong>bowhead</strong> <strong>whale</strong> life history and population dynamics (Rugh et<br />
al., 2007).<br />
The size of the <strong>bowhead</strong> <strong>photo</strong>graphic database has increased to the point where collection of new<br />
<strong>photo</strong>graphs results in a relatively high probability of re-sighting a previously <strong>photo</strong>graphed marked <strong>whale</strong>. The<br />
proportion of the population that has been <strong>photo</strong>graphed has increased, and consequently, the likelihood of<br />
<strong>photo</strong>graphing a <strong>whale</strong> that has been <strong>photo</strong>graphed during previous studies has increased dramatically. Thus,<br />
new <strong>photo</strong>graphs at this time provide much more in<strong>for</strong>mation both on individual <strong>whale</strong>s and the population than<br />
the same number of <strong>photo</strong>graphs did during the early years of <strong>bowhead</strong> <strong>whale</strong> <strong>photo</strong>graphy studies (Schweder<br />
and Sadykova, MS). However, as the number of marked <strong>whale</strong>s in the collection has increased, the time required<br />
to match each new <strong>photo</strong>graph to earlier marked <strong>whale</strong>s has also increased. To assist with the matching of the<br />
rapidly-growing number of <strong>photo</strong>graphs in the BCBS <strong>photo</strong>graphic collection, we have developed a computerassisted<br />
matching program. This paper describes the development and current capabilities of the program and<br />
discusses future testing and development that will improve the efficiency of the program.<br />
METHODS<br />
Programming<br />
All programming was done using the Matlab (The Mathworks, Inc., Natick, MA) development system. This is<br />
the same system that has been used <strong>for</strong> earlier matching programs <strong>for</strong> sea otters and gray <strong>whale</strong>s and has<br />
replaced the Visual Basic/C++ languages used <strong>for</strong> dolphin matching (Hillman et al., 2003; Markowitz et al.,<br />
2003). The program was written using Matlab’s scripting language and graphical user interface (GUI) developer.<br />
The resulting files can be provided to any potential user who has a current Matlab installation. They can also be<br />
compiled by the programmer to produce MS Windows executable files, which run on any Windows XP or Vista<br />
machine. A Matlab runtime file, which the programmer is permitted by Matlab to distribute to end-users, must<br />
also be installed along with the executable file. For this project, the Matlab runtime and the executable files were<br />
1
SC/60/BRG24.<br />
sent to LGL Limited (LGL) and the National Marine Mammal Laboratory (NMML), and personnel there<br />
successfully installed and ran the program.<br />
Entering <strong>images</strong> into the program<br />
Bowhead <strong>whale</strong> <strong>images</strong> (TIF or JPG files) are opened in the matching program; other common file <strong>for</strong>mats could<br />
be accommodated. When an image is opened, the program automatically creates specific subfolders into which<br />
the program will insert the data about the <strong>whale</strong>’s marking pattern that will be computed from the user’s input.<br />
The in<strong>for</strong>mation <strong>for</strong> each image in a folder will be added to these subfolders by the program as the <strong>images</strong> are<br />
entered.<br />
Once an image is open, the user selects “Geometric Trans<strong>for</strong>m” from a menu. The image appears in a new<br />
window next to a provided “standard” image, which is simply a <strong>whale</strong> image of excellent quality used as a<br />
reference image to help map features on the new image to be entered into the system. The user clicks on two<br />
points in the new image, such as the tip of the rostrum and the end of the fluke, and on the corresponding two<br />
points in the standard image. The program scales, rotates, and translates the new image so that all <strong>images</strong> have<br />
the same rotation and scaling. The normalization parameters are saved and are applied to that image each time it<br />
is read by the program.<br />
Some <strong>images</strong> were not sufficiently normalized by this procedure because the <strong>whale</strong> was flexed or twisted in the<br />
image. There<strong>for</strong>e, an “Un-twist” option was added to the image capture procedure. The user selects “Un-twist”<br />
from the menu and then clicks on the tip of the rostrum, a series of at least four points along the anatomic<br />
midline and the base of the fluke. The program per<strong>for</strong>ms a nonlinear distortion of the image that causes the<br />
designated points to fall on the line between the rostrum and the fluke, with interpolation of all pixels between<br />
the selected points. When the control points have been selected, the image is rotated such that the first and last<br />
points fall on the same horizontal row. A spline is then drawn through the intermediate points, with a spline point<br />
computed at each pixel column. Each column is then revolved along its vertical length so that the spline points<br />
all fall on the horizontal between the snout and the tail. The entire image is then rotated back to its original<br />
orientation. This process is fairly slow (at least several seconds), so it should not be used unless there is enough<br />
curvature or twisting to disrupt the location of identifying marks. This process is automatic once the six or more<br />
control points have been selected.<br />
The user then selects the markings on the <strong>whale</strong> that are to be used <strong>for</strong> <strong>identification</strong>. First, a mouse click selects<br />
each mark. If the mark is a small dot, no further action is needed. If it is a large mark or has a distinctive shape,<br />
the user may click on a series of points around the mark to outline it, and the data about its size and shape are<br />
stored <strong>for</strong> use in matching. Users may review an image and add, edit, or delete stored marks without re-doing the<br />
entire mark selection <strong>for</strong> that image. This feature allows updating of mark in<strong>for</strong>mation <strong>for</strong> a <strong>whale</strong> from second<br />
and subsequent <strong>photo</strong>s of the same <strong>whale</strong>. This ensures that the mapped image is an accurate representation of a<br />
<strong>whale</strong> and permits us to incorporate in<strong>for</strong>mation from multiple <strong>images</strong> of a <strong>whale</strong> into a single image <strong>for</strong><br />
matching.<br />
Because some regions of a <strong>whale</strong> are not clearly visible on some <strong>photo</strong>graphs due to wash, glare, etc., the user<br />
may designate all (the program default) or part of each image as a region of interest (ROI) <strong>for</strong> matching. This is<br />
done by clicking around the ROI to create a polygon that includes only the ROI. If such a region is designated,<br />
the program will disregard marks outside of the ROI in other <strong>images</strong> during the matching process. If no ROI is<br />
selected, the entire image is treated as the ROI.<br />
The scaling and orientation, un-twisting, ROI, and locations and descriptions of the distinguishing marks are<br />
automatically stored in the subfolders, and are recalled by the program as needed.<br />
To enter, normalize, and mark an image takes 3-7 minutes, depending on the quality of the image and how many<br />
spots must be marked or outlined. When this process is complete, the user selects “Add to Excel File.” This adds<br />
the file path and name of the new image to a list maintained in a MS Excel file; this file stores the names of all<br />
files in the catalogue and the identity of the <strong>whale</strong> in the image. When the image is first added to the file, the<br />
program adds the identifier “0000” as its ID, later to be changed by the user when the image is identified either<br />
as a new individual or as a match to a previous image. The files listed in the Excel document are the ones to<br />
which each new image is matched.<br />
Matching Images<br />
The user selects any image <strong>for</strong> <strong>identification</strong>. In order to be matched to the rest of the collection, the image must<br />
have been previously normalized and its spots marked. If the normalization and spot marking procedures have<br />
not been completed <strong>for</strong> the selected image, a notice reminds the user to complete them.<br />
2
SC/60/BRG24.<br />
Since, the size, orientation and shape of scars are sometimes distinctive, the user is given an option to indicate<br />
the shape of any mark rather than indicating only its location. For example, this option would be used if a <strong>whale</strong><br />
had a distinctively-shaped mark so that all <strong>whale</strong>s with a similar-shaped mark at the same location would be<br />
ranked higher on the list of potential matches than they might be without selection of this option (Fig. 1). The<br />
parameters measured <strong>for</strong> each mark are: location, area (number of pixels), eccentricity (a numerical indication<br />
of the shape of the spot, ranging from 0 (round) to 1 (straight line)), orientation (angle in degrees from horizontal<br />
of the major axis), length (the major axis, measured in pixels, of an ellipse that approximates the mark) and<br />
perimeter (number of pixels). These parameters are calculated automatically once the user has outlined a spot<br />
with the mouse. If a spot’s location is marked, but the spot is not outlined, the system assumes that the spot is a<br />
small circle. When marks are matched, the difference between them is the Euclidean distance in the data sevendimensional<br />
space of these parameters (x position, y position, area, eccentricity, eccentricity×orientation [i.e., if<br />
it is round its orientation is not relevant], length and perimeter). Each of the parameters is weighted by the<br />
default value unless other values are selected. The default weight is 1 <strong>for</strong> all parameters except eccentricity, <strong>for</strong><br />
which the weight is 10 (because its native value is limited to 0 to 1) and 0.01 <strong>for</strong> area. The user can set the<br />
weighting value to zero to <strong>for</strong>ce a parameter not to contribute to the matching process, or set the weight to any<br />
higher number to have a characteristic of the mark weighted more heavily. The program saves the new weights<br />
in a file called “wtfile.mat.” The capability <strong>for</strong> the user to vary the weights is a part of the development process,<br />
as experience will tell us which of the computed parameters are most helpful <strong>for</strong> <strong>identification</strong>, and there<strong>for</strong>e<br />
should be weighted more heavily, and which, if any, should be counted less heavily or not at all. It is also<br />
possible that the user may wish to adjust the weights according to the properties of the query image<br />
Any two <strong>images</strong> may be compared by reading one as “Image 1” and the other as “Image 2.” A “Compare two<br />
<strong>images</strong>” button measures and displays the difference between the two <strong>images</strong>. This quick matching function is<br />
useful during testing of the program and <strong>for</strong> new users to learn how changing options affects the matching<br />
process.<br />
In the usual operation of the program, the user does not open an “Image 2”, but instead, clicks a “Compare to<br />
List” button. In this case, all of the <strong>images</strong> listed in the Excel file are matched to Image 1. The program measures<br />
the dissimilarity between the query image and each listed image, and displays a list of <strong>images</strong> ordered from the<br />
lowest to the highest dissimilarity value. As a result, the <strong>images</strong> that are most similar to the query image are at<br />
the top of the list. It then displays a window with the sorted list and the dissimilarity value. The program also<br />
displays the top six <strong>images</strong> on the list along with file names and ID’s (Fig. 2). Each of these six <strong>images</strong> has a<br />
button that allows the image to be displayed in a separate large window, with its selected marks shown. If the<br />
user sees a match among those six <strong>images</strong>, he enters the ID of the query image in a provided text box. If a match<br />
is not seen, the user may view the next six. The user can scroll through the catalogue of <strong>images</strong> in either<br />
direction, displaying them six at a time, until a match is found, or the user is convinced that there is no match.<br />
Either a pre-existing or a new ID may be entered, and it is recorded in the Excel file instead of the default<br />
“0000”.<br />
The presence or absence of a match is confirmed by a trained observer with assistance of the program. The goal<br />
of the program is to reduce the number of <strong>images</strong> that must be reviewed. If the program worked perfectly, its<br />
first suggestion would always be a match, if one existed. This ideal goal cannot always be achieved because<br />
some matching errors are inevitable, some <strong>images</strong> are unclear and some <strong>whale</strong>s either lack distinctive markings<br />
or have markings resembling those on other individuals. Nevertheless, if a correct match occurs at or near the top<br />
of the list, the amount of labour previously needed to search <strong>for</strong> matches will be reduced..<br />
Comparison to earlier programs<br />
The overall design of the program is modelled on previous <strong>photo</strong>-ID programs that were developed <strong>for</strong> other<br />
species such as dolphins, sea otters, and gray <strong>whale</strong>s (Hillman et al., 2003; Markowitz et al., 2003). The<br />
<strong>bowhead</strong> program code is entirely new, with many features that are specific to the <strong>bowhead</strong> <strong>whale</strong> application.<br />
The processes of file access, Excel file access, image display and mouse-use take advantage of Matlab functions.<br />
The scale-rotation-translation normalization process uses Matlab’s linear con<strong>for</strong>mal trans<strong>for</strong>mation. However,<br />
the un-twisting function is a new algorithm. This process is fairly slow (at least several seconds), so it should not<br />
be used unless there is enough curvature or twisting to disrupt the location of identifying marks.<br />
Matching function<br />
The matching function works as follows, with the process repeated as the query image is matched in turn to each<br />
image in the collection:<br />
• The program recalls the locations and shape properties of the two <strong>images</strong> (A and B) being matched.<br />
3
SC/60/BRG24.<br />
• If A or B has a designated ROI, the marks in the other image are tested to determine whether they fall<br />
within this region. Any marks that are outside the other image’s valid region are temporarily discarded.<br />
• The Euclidian distance is determined between each point in A and B in a 7-dimensional space<br />
comprised of the X distance, Y distance, mark area, mark eccentricity (a scale from 0 to 1 indicating shape from<br />
a circle to a line), angular orientation of the mark, length of the major axis of the mark and perimeter of the<br />
mark, each parameter weighted by its weighting value. The closest points are determined. If the <strong>images</strong> have<br />
unequal numbers of marks, the smaller number is used (e.g. if A has 3 and B has 5 marks, the closest 3 pairs of<br />
points are found)<br />
• Using a linear translation function and Matlab’s optimization algorithm, the relative position of the two<br />
<strong>images</strong> is adjusted to minimize the distances between the corresponding closest-fit points.<br />
The difference values between the query and all other <strong>images</strong> are accumulated and sorted, establishing the order<br />
in which <strong>images</strong> are presented <strong>for</strong> inspection.<br />
A second matching algorithm was also implemented after discussion with an engineering colleague, Dr. Hemant<br />
Tagare at Yale University. Rather than the successive-approximations optimization of the point locations, the<br />
algorithm uses an analytical minimization of the summed distance, with a specified number of iterations. The<br />
advantage of this method is that it can weight the distances in such a way that the matching value is weighted<br />
more heavily by points that fit more closely than those that are distant matches. The result of this weighting is to<br />
reduce the influence of outlying marks that might be due to errors or artefacts. The user of the program can select<br />
either of the two matching algorithms; experience will tell whether one is actually more effective than the other.<br />
DISCUSSION<br />
A computer-assisted matching program has been developed using 20 <strong>bowhead</strong> <strong>whale</strong> <strong>images</strong> and has been tested<br />
with a set of 78 <strong>images</strong>. These relatively small collections were used to develop the program and determine<br />
whether the programming was free of errors, but we have not used the program with larger numbers of <strong>images</strong>.<br />
A much larger data set is now being prepared <strong>for</strong> testing. The program runs without bugs or crashes, but does not<br />
yet reliably trap user errors or other unexpected input. The program can be improved by adding features such as<br />
internal help files and additional user error trapping. The latest version of the program was provided to LGL and<br />
NMML on 23 April 2008.<br />
A closer examination of weights is needed to determine relative influence of distance, area, etc. of marks on the<br />
similarity of a matched pair of <strong>images</strong>; it is still uncertain what values to assign to these weights in order to<br />
provide the most accurate matching. In order to facilitate the matching process, the program also needs to allow<br />
more flexibility in organizing data files. Most importantly, the program needs to be tested using a large<br />
collection of <strong>images</strong> that have already been matched to determine how far down the list of suggested matches a<br />
correct match may occur. Additional features that may improve the reliability of the matching process have been<br />
requested by the testers at LGL and NMML. Throughout the development process there has been frequent and<br />
effective communication between the programmer and the testing groups, and the result is a working draft of a<br />
program that may be extremely helpful <strong>for</strong> matching <strong>bowhead</strong> <strong>whale</strong> <strong>images</strong>.<br />
ACKNOWLEDGEMENTS<br />
Development and testing of the computer-assisted matching program was by funded Minerals Management<br />
Service (MMS) through an inter-agency agreement between MMS and NMML. We thank Chuck Monnett of<br />
MMS <strong>for</strong> administering the contract and comments on earlier drafts of this paper. We also thank Sally Mizroch<br />
and Kim Goetz of NMML and Rolph Davis of LGL <strong>for</strong> providing comments to improve it.<br />
REFERENCES<br />
Angliss, R.P., Rugh, D.J., Withrow, D.E. and Hobbs, R.C. 1995. Evaluations of aerial <strong>photo</strong>grammetric length<br />
measurements of the Bering-Chukchi-Beau<strong>for</strong>t seas stock of <strong>bowhead</strong> <strong>whale</strong>s (Balaena mysticetus). Rep. int.<br />
Whal. Commn 45:313-24.<br />
da Silva, C.Q., Zeh, J., Madigan, D., Laake, J., Rugh, D., Baraff, L., Koski, W. and Miller, G. 2000. Capturerecapture<br />
estimation of <strong>bowhead</strong> <strong>whale</strong> population size using <strong>photo</strong>-<strong>identification</strong> data. J. Cetacean Res.<br />
Manage. 2(1):45-61.<br />
4
SC/60/BRG24.<br />
Davis, R.A., Koski, W.R. and Miller, G.W. 1983. Preliminary assessment of the length-frequency distribution<br />
and gross annual recruitment rate of the western arctic <strong>bowhead</strong> <strong>whale</strong> as determined with low-level aerial<br />
<strong>photo</strong>grammetry, with comments on life history. Rep. to Natl Mar. Mammal Lab., Natl Mar. Fish. Serv.,<br />
Seattle, WA by LGL Ltd. Paper SC/35/PS5 presented to the IWC Scientific Committee, June 1983<br />
(unpublished). 91pp.<br />
Hillman, G.R., Wuersig, B., Gailey, G.A., Kehtarnavaz, N, Drobyshevsky, A., Araabi, B.N., Tagare, H.D., and<br />
Weller, D.W. 2003. Computer-assisted <strong>photo</strong>-<strong>identification</strong> of individual marine vertebrates: a multi-species<br />
system. Aquatic Mammals 29(1):117-123.<br />
Koski, W.R., Davis, R.A, Miller, G.W., and Withrow, D.E. 1992. Growth rates of <strong>bowhead</strong> <strong>whale</strong>s as<br />
determined from low-level aerial <strong>photo</strong>grammetry. Rep. int. Whal. Commn 42:491-499.<br />
Koski, W.R., Davis, R.A., Miller, G.W. and Withrow, D.E. 1993. Reproduction. pp. 239-74. In: J.J. Burns, J.J.<br />
Montague and C.J. Cowles (eds) The Bowhead Whale. Special Publication. No. 2. Society <strong>for</strong> Marine<br />
Mammalogy, Lawrence, Kansas. 787pp.<br />
Koski, W.R., Rugh, D.J., Punt, A.E. and Zeh, J. 2006. An approach to minimise bias in estimation of the lengthfrequency<br />
distribution of <strong>bowhead</strong> <strong>whale</strong>s (Balaena mysticetus) from aerial <strong>photo</strong>grammetric data. J. Cetacean<br />
Res. Manage. 8:45-54.<br />
Koski, W.R., George, J.C. and Zeh, J. 2007. A calf index <strong>for</strong> monitoring reproductive success in BCB <strong>bowhead</strong><br />
<strong>whale</strong>s (Balaena mysticetus). Paper SC/59/BRG7 presented to the Int. Whal. Commn Scientific Committee,<br />
Anchorage, AK, 7-18 May 2007. 6pp.<br />
Koski, W.R., Mocklin, J., Davis, A.R., Zeh, J., Rugh, D.J., George, J.C and Suydam, R. 2008. Estimates of<br />
Bering-Chukchi-Beau<strong>for</strong>t <strong>bowhead</strong> <strong>whale</strong> (Balaena mysticetus) abundance from 2003-2005 <strong>photo</strong><strong>identification</strong><br />
data. Paper SC/60/BRG18 presented to the Int. Whal. Commn Scientific Committee, Santiago,<br />
Chile, 1-13 June 2008.<br />
Markowitz, T.M, Harlin, A.D, Wuersig, B. 2003. Digital <strong>photo</strong>graphy improves efficiency of individual dolphin<br />
<strong>identification</strong>. Mar. Mamm. Sci. 19(1):217-223.<br />
Miller, G.W., Davis, R.A, Koski, W.R., Crone, M.J., Rugh, D.J., Withrow, D.E. and Fraker, M. 1992. Calving<br />
intervals of <strong>bowhead</strong> <strong>whale</strong>s -- An analysis of <strong>photo</strong>graphic data. Rep. Int. Whal. Commn 42:501-506.<br />
Rugh, D. 1990. Bowhead <strong>whale</strong>s reidentified through aerial <strong>photo</strong>graphy near Point Barrow, Alaska. Rep. int.<br />
Whal. Commn (Spec. Issue 12):289-94.<br />
Rugh, D.J., Miller, G.W., Withrow, D.E. and Koski, W.R. 1992a. Bowhead <strong>whale</strong> calving intervals established<br />
through <strong>photo</strong>graphic re<strong>identification</strong>s. J. Mammal. 73:487-490.<br />
Rugh, D., Koski, W., and George, J. 2007. Interyear re-<strong>identification</strong> of <strong>bowhead</strong> <strong>whale</strong>s during their spring<br />
migration past Barrow, Alaska, 1984-2004. Paper SC/59/BRG2 presented to the IWC Scientific Committee,<br />
Anchorage, AK, 7-18 May 2007. 9pp.<br />
Schweder, T. 2003. Abundance estimation from multiple <strong>photo</strong> surveys: confidence distributions and reduced<br />
likelihoods <strong>for</strong> <strong>bowhead</strong> <strong>whale</strong>s off Alaska. Biometrics 59:974-83.<br />
Schweder, T. and Sadykova, D. 2007. Event history models <strong>for</strong> capture-recapture surveys with passively marked<br />
individuals. Paper SC/59/BRG28 presented to the Int. Whal. Commn Scientific Committee, Anchorage, AK,<br />
May 2007. 14pp.<br />
Schweder, T. and Sadykova, D. MS. Breaking the square root law. (unpublished). 9pp.<br />
Whitcher, B., Zeh, J., Rugh, D., Koski, W. and Miller, G. 1996. Estimation of adult <strong>bowhead</strong> <strong>whale</strong> survival<br />
rates. Paper SC/48/AS12 presented to the IWC Scientific Committee, September 1997 (unpublished). 22pp.<br />
Zeh, J., Poole, D., Miller, G., Koski, W., Baraff, L. and Rugh, D. 2002. Survival of <strong>bowhead</strong> <strong>whale</strong>s, Balaena<br />
mysticetus, estimated from 1981-1998 <strong>photo</strong><strong>identification</strong> data. Biometrics 58:832-40.<br />
5
SC/60/BRG24.<br />
Fig. 1. The value of using the “Weight Attributes” function is illustrated in the program ranking of the above<br />
<strong>images</strong> as matches. The top left panel shows <strong>whale</strong> 1 with spots marked. The top right panel shows <strong>whale</strong> 2 with<br />
spots marked. The bottom panel shows marks of both <strong>whale</strong>s 1 (plus signs) and <strong>whale</strong> 2 (circles) superimposed<br />
on <strong>whale</strong> 2. These two <strong>whale</strong>s are not a match, but without using the Weight Attributes function the program<br />
recognizes these 2 <strong>whale</strong>s as a good match because 5 marked spots match very well. If however, the Weight<br />
Attributes are adjusted to account <strong>for</strong> the large distinct mark #3, the match probability is decreased and more<br />
accurate matching rank is achieved.<br />
6
SC/60/BRG24.<br />
Fig. 2. Search results screen (front) and main screen (behind).<br />
7